Research On The Diversion Of Criminal Trial Procedures Based On Accelerated Failure Model

The diversion of criminal trial procedures can alleviate the contradiction between the large number of cases and the shortage of personnel to a certain extent,and optimize the allocation of judicial resources.However,there may be inconsistent sentencing under different procedures in the judicial system.Whether judicial efficiency and fairness can "coexist" still needs to be verified.Fairness occupies a core position in the value of criminal litigation,and consistency in sentencing and standardization of sentencing are issues that the legal community,and even the entire society,are very concerned about.The diversion of criminal cases should fully consider the importance and particularity of the value relationship between fairness and efficiency in the judicial field,placing "good" before "fast",that is,pursuing judicial efficiency on the premise of ensuring judicial fairness,and achieving a balance between judicial fairness and efficiency at a higher level.Therefore,this article uses two causal effect identification methods under counterfactual frameworks to analyze the impact of procedural separation on sentencing results in theft cases from both a local and global perspective.In the local analysis,the treatment effect of program grouping at a crime amount of 10,000 yuan is estimated using breakpoint regression based on the grouping of the dataset program and the characteristics of sentencing discretion,and a breakpoint regression method based on ensemble learning is proposed to solve the problem of endogenous variables in the model.In the global analysis,propensity score matching method is used to control the influence of covariates on sentencing results.However,due to the imbalance of procedural groups in the theft dataset,the conventional matching method is ineffective due to the small common support interval between the treatment group and the control group.To address this,this article proposes an AFTPSM algorithm,which maps low-crime amount cases in the simple procedure to the higher crime amount interval through the accelerated failure time model,to achieve the effect of shifting the propensity score of the control group towards the treatment group,increase the common support interval to obtain more broadly covered and more similar matching samples,then calculate the overall average treatment effect.The research results show that in the selected dataset of this article,procedural factors will significantly affect the sentence length,and the average sentence length of ordinary procedure cases is about 10.6 months longer than that of simple procedure cases.And in different crime amount intervals,there is significant heterogeneity in treatment effects,and the local treatment effect shows a clear "increase first and decrease to the right" trend as the stolen amount increases.When the crime amount is between ten thousand and twenty thousand yuan,the average sentence length of ordinary procedures is almost 1.8 times that of simple procedures.Therefore,this article believes that there may be a phenomenon of "procedural justice" and "substantive justice" diverging in judicial practice.To address this,it is necessary to further accelerate the construction of the leniency system for confession and punishment,promote the precise sentencing,and restrain the influence of procedural factors on substantive sentences.